ICM cells can be partitioned robustly through transient synchronization using secreted FGF4

The differentiation of ICM cells into epiblast (EPI) and primitive endoderm (PE) is central in embryonic development. It is known that FGF4 signaling is important in this process, but it remains unclear how cells can be correctly partitioned. Here we model the NANOG-GATA6-FGF4 network, and test all 64 logical regulatory combinations for their ability to partition a group of cells. We found that nearly all the logic combinations allowed for correct partitioning, including a minimal network where self-activation of NANOG and GATA6 was inactivated. However such self-activation increased the robustness of the system. Furthermore, the model also captured the reported changes in cell proportions in response to FGF perturbations. This constrains the possible regulatory logic and predicts the presence of an “OR” gate in cell-cell communication. We repeatedly found that FGF4 coordinated the decision in two phases: A convergence and a bifurcation phase. First FGF4 negative feedback drives the cells to a balanced “battle” state where most cells have intermediate levels of both regulators, thus being double positive. Subsequent bifurcation happens at constant FGF4 level. Together our results suggest that the frequently observed state of multipotency during differentiation may be an emergent phenomenon resulting from inter-cellular negative feedbacks.